We've started seeing cases where we want to allocate lots of one-shot
primitives per-frame and the cost of allocating primitives becomes
important in this case since it can start being noticeable in profiles.
The main cost for allocating primitives was the GArray allocation
and appending the attributes to the array. This updates the code to
simply over allocate the primitive storage so we can embed the list
of attributes directly in that allocation.
If the user later sets new attributes and there isn't enough embedded
space then a separate slice allocation for the new attributes is made
but still this should be far less costly than using a GArray as before.
Most of the time we would expect when setting new attributes there will
still be the same number of attributes, so the embedded space can simple
be reused.
Reviewed-by: Neil Roberts <neil@linux.intel.com>
Calling g_strdup for attribute names was starting to show up in profiles
due to calling malloc for new string storage so frequently. This avoids
calling g_strdup and calls g_intern_string() instead. For the really
common case names we even avoid the cost of g_intern_string since we
can trivially relate our internal name_id to a static string.
Reviewed-by: Neil Roberts <neil@linux.intel.com>
As part of the on going, incremental effort to purge the non type safe
CoglHandle type from the Cogl API this patch tackles most of the
CoglHandle uses relating to textures.
We'd postponed making this change for quite a while because we wanted to
have a clearer understanding of how we wanted to evolve the texture APIs
towards Cogl 2.0 before exposing type safety here which would be
difficult to change later since it would imply breaking APIs.
The basic idea that we are steering towards now is that CoglTexture
can be considered to be the most primitive interface we have for any
object representing a texture. The texture interface would provide
roughly these methods:
cogl_texture_get_width
cogl_texture_get_height
cogl_texture_can_repeat
cogl_texture_can_mipmap
cogl_texture_generate_mipmap;
cogl_texture_get_format
cogl_texture_set_region
cogl_texture_get_region
Besides the texture interface we will then start to expose types
corresponding to specific texture types: CoglTexture2D,
CoglTexture3D, CoglTexture2DSliced, CoglSubTexture, CoglAtlasTexture and
CoglTexturePixmapX11.
We will then also expose an interface for the high-level texture types
we have (such as CoglTexture2DSlice, CoglSubTexture and
CoglAtlasTexture) called CoglMetaTexture. CoglMetaTexture is an
additional interface that lets you iterate a virtual region of a meta
texture and get mappings of primitive textures to sub-regions of that
virtual region. Internally we already have this kind of abstraction for
dealing with sliced texture, sub-textures and atlas textures in a
consistent way, so this will just make that abstraction public. The aim
here is to clarify that there is a difference between primitive textures
(CoglTexture2D/3D) and some of the other high-level textures, and also
enable developers to implement primitives that can support meta textures
since they can only be used with the cogl_rectangle API currently.
The thing that's not so clean-cut with this are the texture constructors
we have currently; such as cogl_texture_new_from_file which no longer
make sense when CoglTexture is considered to be an interface. These
will basically just become convenient factory functions and it's just a
bit unusual that they are within the cogl_texture namespace. It's worth
noting here that all the texture type APIs will also have their own type
specific constructors so these functions will only be used for the
convenience of being able to create a texture without really wanting to
know the details of what type of texture you need. Longer term for 2.0
we may come up with replacement names for these factory functions or the
other thing we are considering is designing some asynchronous factory
functions instead since it's so often detrimental to application
performance to be blocked waiting for a texture to be uploaded to the
GPU.
Reviewed-by: Neil Roberts <neil@linux.intel.com>
Previously the instructions were telling the developer to run
./build/config.guess to get the build name to pass to
configure. However that file only exists after running automake so
it's a bit awkward. This patch makes it download config.guess from the
gitweb for automake and just explicitly run it.
Reviewed-by: Robert Bragg <robert@linux.intel.com>
The -c option for wget and -C - option to curl are used to make it
continue the download if the file already exists. The idea was that it
wouldn't waste time downloading the files again if the file already
exists. However this causes problems if the remote file gets larger
because the download will continue from the size of the old file so it
will get corrupt. Instead let's just explicitly check if the file
already exists and avoid calling wget or curl altogether.
Reviewed-by: Robert Bragg <robert@linux.intel.com>
mesa_wgl.h can no longer be fetched from upstream and since it's no
longer used anyway we don't fetch this any more. This also updates
the blurb printed after fetching dependencies to show how to run
./configure so we pass --enable-wgl not --enable-stub-winsys and
to also pass the -I path for the cogl-cross/include directory which has
the latest gl.h we fetched so the build doesn't try and use the headers
shipped with the mingw toolchain which may be out-of-date.
Reviewed-by: Neil Roberts <neil@linux.intel.com>
When testing with COGL_DEBUG=disable-npot-textures all of the tests
would fail because the testing infrastructure itself ends up creating
a sliced texture and then trying to use it as a render target. This
just modifies test-utils to use 512x512 for the size of the texture.
Reviewed-by: Robert Bragg <robert@linux.intel.com>
cogl_rectangle has some validation code to check whether the first
layer has a sliced texture. If so it will abandon the rest of the
layers and print a warning. However it was even doing this pruning and
displaying the warning if there is only one layer. This patch just
makes it check whether the pipeline actually has more than one layer
before pruning or displaying the warning but it will still fallback to
the multiple quads path.
Reviewed-by: Robert Bragg <robert@linux.intel.com>
This adds an internal function to set the backface culling state on a
pipeline. This includes properties to set the culling mode (front,
back or both) and also to set which face is considered the front
(COGL_WINDING_CLOCKWISE or COGL_WINDING_COUNTER_CLOCKWISE). The actual
front face flushed to GL depends on whether we are rendering to an
offscreen buffer or not. This means that when changing between on- and
off- screen framebuffers it now checks whether the last flushed
pipeline has backface culling enabled and forces a reflush of the cull
face state if so.
The backface culling is now set on a pipeline as part of the legacy
state. This is important because some code in Cogl assumes it can
flush a temporary pipeline to revert to a known state, but previously
this wouldn't disable backface culling so things such as flushing the
clip stack could get confused.
Reviewed-by: Robert Bragg <robert@linux.intel.com>
When changing between two framebuffers that have different color masks
it now forces the pipeline to flush the mask by setting
current_pipeline_changes_since_flush. For this to work there needs to
be a common bit of code that gets called when the framebuffers are
changed that has access to both the old framebuffer and the new
framebuffer. _cogl_set_framebuffers_real can't be used for this
because when it is called from cogl_pop_framebuffer the stack entries
have already changed so it can't know the old framebuffer. This patch
adds a new function called notify_buffers_changed which should get
called whenever the buffers are changed and it explicitly gets passed
pointers to the old and new buffers. cogl_pop_framebuffer now calls
this instead of trying to use _cogl_set_framebuffers_real to force a
flush.
This patch also fixes the ctx->window_buffer pointer. Previously this
was implemented by searching in the framebuffer stack for an onscreen
framebuffer whenever the current buffers are changed. However it does
this after the stack has already changed so it won't usually find the
right buffer.
Reviewed-by: Robert Bragg <robert@linux.intel.com>
There is a currently a bug where pushing a buffer with a different
color mask will not cause the color mask to be flushed. This adds a
test to demonstrate that.
Reviewed-by: Robert Bragg <robert@linux.intel.com>
The ARBfp backend can't handle fog so it tries to check for when it's
enabled and bails out. However it was checking using the global legacy
state value on the CoglContext but this doesn't necessarily reflect
the state that will actually be used by the pipeline because Cogl may
have internally pushed a different pipeline.
This patch adds an internal _cogl_pipeline_get_fog_enabled which the
ARBfp backend now uses.
Reviewed-by: Robert Bragg <robert@linux.intel.com>
Some code in Cogl such as when flushing a stencil clip assumes that it
can push a temporary simple pipeline to reset to a known state for
internal drawing operations. However this breaks down if the
application has set any legacy state because that is set globally so
it will also get applied to the internal pipeline.
_cogl_draw_attributes already had an internal flag to disable applying
the legacy state but I think this is quite awkward to use because not
all places that push a pipeline draw the attribute buffers directly so
it is difficult to pass the flag down through the layers.
Conceptually the legacy state is meant to be like a layer on top of
the purely pipeline-based state API so I think ideally we should have
an internal function to push the source without the applying the
legacy state. The legacy state can't be applied as the pipeline is
pushed because the global state can be modified even after it is
pushed. This patch adds a _cogl_push_source() function which takes an
extra boolean flag to mark whether to enable the legacy state. The
value of this flag is stored alongside the pipeline in the pipeline
stack. Another new internal function called
_cogl_get_enable_legacy_state queries whether the top entry in the
pipeline stack has legacy state enabled. cogl-primitives and the
vertex array drawing code now use this to determine whether to apply
the legacy state when drawing. The COGL_DRAW_SKIP_LEGACY_STATE flag is
now removed.
Reviewed-by: Robert Bragg <robert@linux.intel.com>
Since 12b3d21aaa cogl is using the vertex attribute API to stroke a
path. However it was still manually appllying the legacy state to the
pipeline. cogl_vdraw_attributes also applies the legacy state so it
ends up getting applied twice. This patch just removes it from
_cogl_path_stroke_nodes.
Reviewed-by: Robert Bragg <robert@linux.intel.com>
This patch basically restores the logic from 1.6. There we assumed that
glXCopySubBuffer won't tear and thus only needs to be throttled to the
framerate, while glBlitFramebuffer needs to always wait to avoid
tearing.
With Nvidia drivers specifically we have seen that glBlitFramebuffer is
not synchronized. Eventually the plan is that Cogl will actually take
into consideration the underlying driver/hw vendor and driver version
and we may want to only mark glBlitFramebuffer un-synchronized on
Nvidia.
https://bugzilla.gnome.org/show_bug.cgi?id=659360
Reviewed-by: Robert Bragg <robert@linux.intel.com>
All of the cogl_pipeline API is currently experimental so this makes
sure the API is surrounded by #ifdef COGL_ENABLE_EXPERIMENTAL_API
guards and all the symbols have a #define to give them an _EXP suffix as
we do for other experimental API.
Reviewed-by: Neil Roberts <neil@linux.intel.com>
As part of an on-going effort to get cogl-pipeline.c into a more
maintainable state this splits out all the apis relating just to
layer state. This just leaves code relating to the core CoglPipeline
and CoglPipelineLayer design left in cogl-pipeline.c.
This splits out around 2k more lines from cogl-pipeline.c although we
are still left with nearly 4k lines so we still have some way to go!
Reviewed-by: Neil Roberts <neil@linux.intel.com>
Since cogl-pipeline.c has become very unwieldy this make a start at
trying to shape this code back into a manageable state. This patche
moves all the API relating to core pipeline state into
cogl-pipeline-state.c. This doesn't move code relating to layer state
out nor does it move any of the code supporting the core design
of CoglPipeline itself.
This change alone factors out 2k lines of code from cogl-pipeline.c
which is obviously a good start. The next step will be to factor
out the layer state and then probably look at breaking all of this
state code down into state-groups.
Reviewed-by: Neil Roberts <neil@linux.intel.com>
dgettext (which Cogl is using) doesn't work unless you first tell
gettext where the locale dir is for the library's domain. This just
adds the necessary calls into _cogl_init.
https://bugzilla.gnome.org/show_bug.cgi?id=658700
Reviewed-by: Robert Bragg <robert@linux.intel.com>
When the clip contains two rectangles which do not intersect it was
generating a clip bounds where the bottom-right corner was above or to
the left of the top-left corner. This would end up allowing the pixels
between the two rectangles instead of clipping everything like it
should. To fix this there is now an extra check which detects this
situation and just clears the clip bounds to all zeroes in a similar
way to what cogl-clip-stack does.
https://bugzilla.gnome.org/show_bug.cgi?id=659029
Reviewed-by: Robert Bragg <robert@linux.intel.com>
Commit 12b3d21a changed cogl-path so that it will use the vertex
attribute API to stroke the path in a similar way to how it was using
the API to fill the path. However it wasn't clearing the stroke buffer
when the path is modified so it would continue to use the unmodified
stroke.
Reviewed-by: Robert Bragg <robert@linux.intel.com>
Instead of adding -DCOGL_ENABLE_EXPERIMENTAL_2_0_API to the
cogl-2.0-experimental.pc file we now install a cogl2-experimental.h
that #defines COGL_ENABLE_EXPERIMENTAL_2_0_API before including
cogl.h.
The problem with having the define in the .pc file is that you might
develop a library that depends on the experimental 2.0 api internally
and then you might want to use that library with Clutter which still
uses the 1.0 API but the .pc file for your library will indirectly,
automatically enable the 2.0 api which can cause conflicts.
When we are about to start arbfp codegen we call shader_state_new() to
allocate new state structures used to build up the code and that
function makes sure to zero the newly allocated structures.
Right after calling shader_state_new() we were then also explicitly
iterating though the newly allocated unit_state structures and zeroing
the .sampled and .dirty_combine_constant members as well as resetting
shader_state->next_constant_id = 0. This patch removes that redundant
re-initialization of state.
Reviewed-by: Neil Roberts <neil@linux.intel.com>
We weren't actually tracking which layers have been allocated param
space for combine constants; all layers just had a default constant_id
of 0 that indexes into the program.local[] params array and a dirty flag
to say when the constant needs updating. There are times though when we
say to update everything by-passing the dirty flag and because we
weren't actually tracking which layers needed constants we would always
write a constant to program.local[0] for every layer. The upshot was
that we could end up clobbering a real constant that was actually
allocated the constant_id = 0 slot.
This patch adds a new UnitState bitfield to track if the layer has a
corresponding constant that may need flushing and we only ever write the
constant with glProgramLocalParameter4fv if that's set.
https://bugzilla.gnome.org/show_bug.cgi?id=658092
Reviewed-by: Neil Roberts <neil@linux.intel.com>
